Update CI

[orbeckst]

Fixes https://github.com/MDAnalysis/mdaencore/security/dependabot/2

Changes made in this Pull Request:

• update old/deprecated actions
• update tested Python to 3.11 - 3.14
• update deployment workflows
• update RTD config

PR Checklist

• Tests?
• Docs?
• CHANGELOG updated?
• Issue raised/referenced?

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.19%. Comparing base (0ebcaff) to head (029ae45).
⚠️ Report is 2 commits behind head on main.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[orbeckst]

The windows runners fail when re-installing MDAnalysis. It looks as if the mamba line just doesn't do anything:

Install line: MDAnalysis MDAnalysisTests
Run mamba install -y -c conda-forge MDAnalysis MDAnalysisTests
  mamba install -y -c conda-forge MDAnalysis MDAnalysisTests
  shell: C:\Program Files\Git\bin\bash.EXE -l {0}
  env:
    INPUT_RUN_POST: true
    CONDA: C:\Users\runneradmin\miniconda3
Microsoft Windows [Version 10.0.26100.32522]
(c) Microsoft Corporation. All rights reserved.

D:\a\mdaencore\mdaencore>

The above should install something, right?

When we then try to import

Run pyver=$(python -c 'import MDAnalysis; print(MDAnalysis.__version__)')
Traceback (most recent call last):
  File "<string>", line 1, in <module>
    import MDAnalysis; print(MDAnalysis.__version__)
    ^^^^^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'MDAnalysis'
Error: Process completed with exit code 1.

then there's no MDAnalysis package.

[orbeckst]

The windows/mamba failure is fixed by setting installer: conda — this seems to work reliably across all runners. See also MDAnalysis/cookiecutter-mdakit#140

[orbeckst]

Deploy testing

• testpypi deployment and testing
• pypi

[orbeckst]

Currently, deployment requires explicit approval, which slows things down a bit and I am not sure if the extra security is needed. I am not a maintainer here so I am not going to change that but wanted to at least let @MDAnalysis/encore folks know.

[orbeckst]

@enoee @yuxuanzhuang this PR updates the CI so that both tests and deployments work again.

It also raises the supported Python versions to 3.11-3.14. Once this is merged, we can look at #72 and the PR #78 to see if they are still needed. Either way, we can then make a 1.0.1 release that should hopefully work again on all platforms and on all recent Pythons.

Please have a quick look. Once you're happy, I'll squash-merge.

[enoee]

Thanks a lot for doing this, I have no comments

[orbeckst]

The current deployment had an inconsistency because I had not understood how pypa/cibuildwheel was driven. In order to produce arm64 wheels we should (probably) not use macos15-intel but macos-15 (see the images for public runners).

For comparison, it was interesting to see how scikit-learn builds wheels.

[ojeda-e]

I'm shadowing your changes here.
Since I’m not very familiar with this action (or conda in general), I wanted to ask: is this python-version meant to target the latest Python version? I’m wondering whether it might be better to target the oldest supported version instead, then if it works with the min supported here -3.11-, it should also work for newer versions.

[orbeckst]

I don't think it matters as it really only needs to build the docs. I just bumped it up so that it will be 2 years until someone has to bump it up again...

[IAlibay]

Currently, deployment requires explicit approval, which slows things down a bit and I am not sure if the extra security is needed. I am not a maintainer here so I am not going to change that but wanted to at least let @MDAnalysis/encore folks know.

Generally the recommendation is to keep the approval step - it's what we do in the core library. You don't want anyone attempting to trigger a release without maintainer approval.

[orbeckst]

Ok, thanks for the advice. I suppose anyone can git-tag and that triggers the test-pypi deployment, which is then stopped through the approval action (or is git-tag functionality restricted?)

[orbeckst]

Leave the branch alive. I used 1.0.1rc[12] tags during testing and these tags (and releases) point to commits on this branch. We should probably eventually remove all of them (including testpypi and pypi)....